Hemodynamic Characterization of Peripheral Arterio-Venous Malformations Using Rapid Contrast-Enhanced MR Imaging: An In Vitro and In Vivo Study.

Purpose Peripheral arterio-venous malformations (pAVMs) are vascular defects often requiring extensive medical treatment. To improve disease management, hemodynamic markers based on 2D Digital Subtraction Angiography (DSA) data were previously defined to classify pAVMs. However, DSA offers only 2D information, involves ionizing radiation, and requires intra-arterial intervention. We hypothesized that pAVMs could be classified with the same approach with 3D dynamic contrast-enhanced MR-based data. To this end, the present work aims to develop a computational classification system for pAVMs using 3D dynamic contrast-enhanced MR-based data. Methods A pAVM phantom was imaged using both DSA and MRI to validate the methodology, which was then applied to 10 MR-based in vivo datasets. A semi-automated vessel detection algorithm, based on the standard deviation of each voxel or pixel in time, was used. Classification was performed by identifying the time of arrival (CAToA) of contrast agent (CA) and the maximum time derivative of the CA transport in each pixel or voxel (CAsi). Results Normalized CAToA and CAsi histograms showed no significant difference between in vitro DSA and MRI (respectively χ2 = 0.20, p = 0.65 and χ2 = 0.21, p = 0.65), validating the methodology to classify pAVMs. CAToA histograms for type II-IV AVMs derived from in vivo MR-based data aligned with DSA patterns and known hemodynamics. CAToA histograms of capillary-venulous AVMs were distinct, with non-zero values at later times than other AVM types, representing late venous drainage. Type IV AVMs histograms for CAsi were more right-skewed than those derived from types II and III pAVMs. Conclusions MR image quality and temporal resolution are sufficient to allow a classification of pAVMs. This classification method has the potential to become a diagnostic tool for the surgical navigation of pAVMs for clinicians.